This post was written by Sander Rodenhuis and Posted on 5 maart 2017

A week ago AWS suffered a major S3 service disruption in its US-EAST-1 region. What struck me this week was the amount of posts (mainly hardware vendors) saying this is exactly the reason why you should go for hybrid. Is this true, or did the affected customers just didn’t design for failure?

Okay, the S3 service disruption in its US-EAST-1 region was to say the least not a good advertisement for public cloud offerings. But to my opinion it has nothing to do with a hybrid or multi cloud vendor strategy. In time everything will fail at some point. If you need high availability, you should design for it.

In 2015 AWS released S3 cross-region replication. With cross-region replication every object uploaded to a particular S3 bucket is automatically replicated to a designated destination bucket located in a different AWS region. Just enable versioning, select enable cross-region replication and select the destination region. So, that wasn’t that difficult right? Okay, there is a down side: transferring data to another region will cost more (I don’t have the correct figures present, but I think it’s somewhere around 70%). But hey, storing 1TB of data in a S3 Standard storage bucket will cost you around EUR 44,60 a month! Big deal. Sure, you’ll need more than storage alone, but it is possible to automate a complete deployment of your application in another region. And you’ll only start paying when the shit hits the fan.

Most on-premise datacenters are not completely resilient to common hardware and software failures and are not capable to support various automated failover patterns. With AWS you can. But you have to know how. What I think we can learn from the S3 service disruption is that most organizations do not have the skills to fully take advantage of cloud services like the ones AWS offers to build completely automated, high available and resilient systems.

Yes, there might be some cases that still require you to run workloads in your on-premise datacenter (although I think you should consider to get rid of them as soon as possible), but to say that you need a hybrid strategy because of the possible unavailability of public cloud services? No way!

A multiple cloud vendor strategy then? Would you run vSphere in the primary datacenter and Hyper-V in the secondary? Then you’ll also need to be able to operate both! And what about interoperability? Every public cloud vendor has its own characteristics. Being able to take full advantage of only one, is – as proven with the recent S3 outage – for most companies already a great challenge.